Yes, I agree -- if we ever give up this decision making to ai models not only will their actions sometimes be misaligned, but there will be no one accountable.

But all we did was ask the AI for world peace. How could that possibly go wrong?